Open access Original research J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from Shed antigen-induced­ blocking effect on CAR-T­ cells targeting Glypican-3 in Hepatocellular Carcinoma

Luan Sun,1 Fang Gao,1 Zhanhui Gao,1,2 Lei Ao,1 Na Li,1 Sujuan Ma,1 Meng Jia,3,4 Nan Li,5 Peihua Lu,6 Beicheng Sun,7 Mitchell Ho,5 Shaochang Jia,4 Tong Ding,1 1 Wei Gao ‍ ‍

To cite: Sun L, Gao F, Gao Z, ABSTRACT including sorafenib and lenvatinib; most et al. Shed antigen-­induced Background Glypican-3 (GPC3), a cell surface HCC patients cannot receive curative treat- blocking effect on CAR-­T glycoprotein that is pathologically highly expressed in cells targeting Glypican-3 in ment based on current restrictive criteria and hepatocellular carcinoma (HCC), is an attractive target for 2 3 Hepatocellular Carcinoma. exhibit an extremely poor prognosis. Addi- immunotherapies, including chimeric antigen receptor Journal for ImmunoTherapy tionally, immunotherapies using immune (CAR) T cells. The serum GPC3 is frequently elevated in of Cancer 2021;9:e001875. checkpoint inhibitors have shown benefits doi:10.1136/jitc-2020-001875 HCC patients due to the shedding effect of cell surface GPC3. The shed GPC3 (sGPC3) is reported to block the in only a small proportion of HCC patients, LS and FG contributed equally. function of cell-surface­ GPC3 as a negative regulator. possibly because of the complex immunosup- 4 5 Accepted 09 March 2021 Therefore, it would be worth investigating the potential pressive microenvironment. Thus, there is influence of antigen shedding in anti-GPC3­ CAR-T­ therapy an urgent need to develop new effective ther- for HCC. apeutic strategies for HCC. Methods In this study, we constructed two types of CAR-T­ Glypican-3 (GPC3) is a cell-­surface glyco- cells targeting distinct epitopes of GPC3 to examine how phosphatidylinositol (GPI)-­anchored protein sGPC3 influences the activation and cytotoxicity of CAR-­T that belongs to the heparan sulfate (HS) cells in vitro and in vivo by introducing sGPC3 positive proteoglycan family, which plays important patient serum or recombinant sGPC3 proteins into HCC cells or by using sGPC3-­overexpressing HCC cell lines. roles in cell growth, differentiation and migra- Results Both humanized YP7 CAR-­T cells and 32A9 tion. Many studies have shown that GPC3 is CAR-T­ cells showed GPC3-­specific antitumor functions highly expressed in HCC, while its expression

in vitro and in vivo. The existence of sGPC3 significantly is absent in most nonmalignant adult tissues. http://jitc.bmj.com/ inhibited the release of cytokines and the cytotoxicity of GPC3 is currently used as an informative anti-GPC3­ CAR-T­ cells in vitro. In animal models, mice immunohistochemical biomarker for HCC, carrying Hep3B xenograft tumors expressing sGPC3 and it is believed to be an attractive target exhibited a worse response to the treatment with CAR-­T for HCC therapy.6 7 Various GPC3-­targeted cells under both a low and high tumor burden. sGPC3 strategies have been developed or evaluated bound to CAR-T­ cells but failed to induce the effective 8 9 activation of CAR-­T cells. Therefore, sGPC3 acted as in HCC. Our previous work showed that on September 30, 2021 by guest. Protected copyright. dominant negative regulators when competed with cell GPC3-­specific immunotoxin and antibody surface GPC3 to bind anti-GPC3­ CAR-T­ cells, leading to an drug conjugates presented potent anticancer 10 11 inhibitory effect on CAR-­T cells in HCC. activity in vitro and in vivo. Recently, Conclusions We provide a proof-of-­ concept­ study chimeric antigen receptor (CAR) T cells, © Author(s) (or their demonstrating that GPC3 shedding might cause worse which have shown promising curative effects employer(s)) 2021. Re-­use response to CAR-T­ cell treatment by competing with cell in hematological tumors, have been applied permitted under CC BY-­NC. No surface GPC3 for CAR-­T cell binding, which revealed a new for the development of novel GPC3-targeted­ commercial re-­use. See rights mechanism of tumor immune escape in HCC, providing therapies in HCC and have shown prelimi- and permissions. Published by a novel biomarker for patient enrolment in future clinical BMJ. nary effects in some studies12–15 and clinical trials and/or treatments with GPC3-­targeted CAR-­T cells. 8 9 16 For numbered affiliations see trials. However, the response rate in the end of article. clinical setting is still far from satisfactory, and it is not clear how HCC cells may escape from Correspondence to BACKGROUND Dr Wei Gao; gao@​ ​njmu.edu.​ ​cn Liver cancer is a leading cause of cancer-­ CAR-­T cells. related death with an increasing incidence GPC3 is composed of a 70 kDa core protein Dr Tong Ding; worldwide. Primary hepatocellular carcinoma and two HS side chains with highly negative dingtong@​ ​njmu.edu.​ ​cn (HCC) is the most common type of liver charges, which can help GPC3 recruit and 1 Dr Shaochang Jia; cancer. HCC is generally resistant to chemo- concentrate many essential ligands from the shaochangjiabio@​ ​163.com​ therapy, radiotherapy and specific inhibitors, tumor microenvironment and facilitate their

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 1 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from recognition of corresponding receptors.17–19 Our previous donors’ blood, obtained under the ethics protocol studies have demonstrated that GPC3 functions as a core- approved by Nanjing Jinling Hospital (Nanjing, Jiangsu). ceptor to modulate Wnt/β-catenin signaling to promote The isolated PBMCs were cultured in RPMI 1640 Medium cell proliferation in HCC.19–21 Similar to other glypicans, (Roswell Park Memorial Institute 1640 Medium, HyClone, GPC3 can be released from the cell surface and can be Logan, Utah, USA) supplemented with 10% fetal bovine found around tumors or in the circulation.22 Several serum (Gibco, Murphysboro, IL), 100 U/mL penicillin, studies have shown that serum GPC3 levels are signifi- and 0.1 mg/mL streptomycin, with incubation under 5% cantly elevated in HCC patients compared with healthy CO2 at 37°C. All other cell lines were cultured in Dulbec- individuals23–25 and that local tumor GPC3 concentrations co’s modified Eagle’s medium (HyClone) supplemented in HCC may also be much higher than those in normal with 10% fetal bovine serum (VACCA, Murphysboro, Illi- tissues. Previous studies also showed that recombinant nois, USA), 100 U/mL penicillin and 0.1 mg/mL strep- soluble GPC3 could inhibit HCC cell growth in vitro tomycin (HyClone). All cell lines were confirmed free of and in vivo,26 27 suggesting that shed GPC3 (sGPC3) may mycoplasma contamination by PCR. compete with cell-­surface GPC3 to bind GPC3-­interacting molecules or may even block GPC3-­targeted therapies. Antibodies and recombinant proteins Many tumor antigens are actively shed from cell YP7 and 32A9 are GPC3-­specific monoclonal antibodies 30–32 surface, and such shedding can be expected to signifi- generated in our previous studies. YP7 is a mouse cantly influence the efficiency of anticancer treatment antibody and was humanized (named hYP7) for the targeting these antigens.28 29 However, it is not very clear generation of CAR-T­ cells. The recombinant sGPC3 how antigen shedding is involved in CAR-T­ therapy, espe- (Q25-­S550) protein with a hexa-histidine­ tag at the cially in solid tumors, which compared with hematolog- C-­terminal end, or with a wild-­type human Fc (hFc) tag ical malignancies exhibit a relatively isolated and enclosed or mutant hFc tag containing three mutations (N297A/ microenvironment. Therefore, it is worth analyzing the L234A/L235A) reported to abolish unexpected binding 33 34 role of sGPC3 in CAR-T­ cell therapy regimens for HCC, to the Fcγ receptor were cloned into the pFUSE with the expectation of identifying novel diagnostic and/ vector (InvivoGen, San Diego, California, USA). All the or prognostic indications. above plasmids were expressed in HEK293T cells. After In the current study, based on GPC3-specific­ antibodies collecting the supernatant, protein purification was 35 developed in our previous work, we established GPC3-­ accomplished following our previous protocol. Recom- targeted CAR-­T cells and studied how sGPC3 influenced binant sCD19(P20-­K291) with either Fc tag or mutant Fc CAR-T­ cell activation and cytotoxicity in HCC models in tag were also purified as the negative control antigen. vitro and in vivo. In short, our results showed that sGPC3 Serum GPC3 detection can impair the antitumor activities of CAR-T­ cells in HCC A 32A9 IgG (4 µg/mL) was used to coat ELISA wells at by competing with cell-sur­ face GPC3, resulting in the 4°C overnight. After blocking, diluted serum or standard escape of tumors from immunotherapy. http://jitc.bmj.com/ recombinant GPC3 protein were added and incubated at 4°C overnight. After washing, biotin labeled YP7-IgG­ (1 µg/mL) was added and then detected by Streptavi- METHODS din-­HRP antibody (Life Tech, Peoria, Illinois, USA). TMB Clinical samples and cells and H SO were added to detect the OD value. The HCC tumor tissue, normal liver tissue specimens 2 4 450nm and serum used in the current study were obtained from Generation and expansion of CAR-T cells on September 30, 2021 by guest. Protected copyright. patients recruited from Nanjing Drum Tower Hospital hYP7, 32A9 and FMC63-based­ second-­generation 4-­1BB (Nanjing, Jiangsu, China) between January 2018 and CARs were designed and packaged into a lentivirus as we April 2019, and Nanjing Jinling Hospital (Nanjing, described previously.32 Then, T cells were isolated from Jiangsu, China) between October 2020 and December PBMCs using a CD8+CD4+ negative isolation kit (STEM- 2020. This study was approved by the Ethical Committee CELL Technologies, Vancouver, Canada) and stimulated of the participating hospitals, and every patient provided with anti-CD3/anti-­ ­CD28 beads (Thermo Fisher Scien- written informed consent. HepG2, Hep3B, A431 and tific, Waltham, Massachusetts, USA) at a bead-­to-­cell HEK293T cells were purchased from the American Type ratio of 2:1 with recombinant human IL-2 (300 U/mL) Culture Collection (Manassas, Virginia, USA). Huh7 cells for 24 hours. T cells were then transduced with the lenti- were obtained from Xin-W­ ei Wang at the National Cancer viruses expressing CARs at a multiplicity of infection of Institute, NIH. The HEK293 TOPFlash stable cell line was 10:1. The transduced T cells were fed with fresh medium a gift from Dr. Jeremy Nathans, Johns Hopkins Medical containing IL-2 and counted every other day. Once the School. The A431-­GPC3 stable cell line was generated in cells were greatly expanded approximately 10 days, they 21 our previous work. The Hep3B-­sGPC3 and Huh7-­sGPC3 were collected for assays or frozen for later use. stable cell lines were generated by transducing cells with a truncated GPC3 (Q25-­S550) sequence in a lentivirus. In vitro CAR-T cell activation and cytotoxicity assays Peripheral blood mononuclear cells (PBMCs) were CAR-­T cells were treated with recombinant sGPC3 or the isolated by the Ficoll density centrifugation of healthy cell culture supernatant or were cocultured with target

2 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from tumor cells (in 96-well­ U-­bottom plates) at different ratios GPC3-knockout­ Hep3B cells exhibited delayed initia- for the indicated time periods. The supernatants were tion and slower growth of xenograft tumors (figure 1B), collected for cytokine release assays with sandwich ELISA verifying the proliferative effect of GPC3 on HCC tumor kits (MultiSciences Biotech, Hangzhou, China) or were progression. used to measure specific cytotoxicity in a lactate dehydro- The mature GPC3 contains a conformational protein genase cytotoxicity assay (Promega, Madison, Wisconsin, core and a C-terminal­ flexible linker. The conforma- USA). tional protein core of GPC3 is made of several α-heliex and strengthened by seven disulfide bounds. The flexible In vivo animal studies C-­terminal linker connected the protein core of GPC3 to Immunodeficient NCG (Nobese Diabetic (NOD)/ cell membrane through GPI linkage. Our previous work ShiLtJGpt-­Prkdcem26Cd52Il2rgem26Cd22/Gpt) mice which were generated two monoclonal antibodies 32A9 and YP7, which established by knocking out Prkdc and Il2rg, and BALB/c targeted the protein core and C-terminal­ flexible region, nu/nu nude mice were purchased from the Model Animal respectively.30 32 Therefore, the GPC3 molecule recognized Research Center of Nanjing University. A total of 5×106 by both 32A9 and YP7 might represent a structure close to luciferase-expressing­ Hep3B, Hep3B-Mock­ or Hep3B-­ its native conformation (figure 1C). Both YP7 and 32A9 sGPC3 cells were suspended in 200 µL of PBS and subcu- recognized cell membrane GPC3 (mGPC3) effectively taneously (s.c.) inoculated into the right flank of 4- to (figure 1D), indicating their potential for application in 6-week-­ old­ NCG female mice that were shaved with an elec- CAR-­T cell construction. We then synthesized 4-1BB­ CAR in tric clipper. Tumor size and weight were measured regu- a second-generation­ format based on hYP7 and 32A9 scFv larly with a digital caliper. Tumor volume was calculated as sequences (figure 1E) and produced virus particles in a length×width2×0.5. When the average tumor size reached lentivirus packaging system. Human T cells were efficiently the indicated size, the mice were intravenously injected transduced with the lentivirus encoding hYP7-­CAR, 32A9-­ with 5–10 million of the indicated T cells in a volume of CAR and control CAR against CD19 (figure 1F and online 200 µL. Tumor lesions were also visualized weekly using the supplemental figure S2). Xenogen IVIS Lumina system (PerkinElmer). Mice were euthanized when they showed any sign of sickness or when CAR-T cells targeting GPC3 inhibited HCC tumor growth in the tumor size reached 2000 mm3. vitro and in vivo To evaluate the antitumor activity of CAR-T­ cells against Statistical analysis GPC3 in vitro, hYP7 CAR-­T cells and 32A9 CAR-­T cells were All statistical analyses were performed using GraphPad incubated with GPC3-­overexpressing A431 cells at different Prism V.8 software (GraphPad Software). Representative ratios for cytotoxicity determination. A431-GPC3­ cells were results were obtained in at least three independent exper- efficiently lysed by the CAR-­T cells targeting GPC3 in a dose-­ iments. The statistical analysis of differences between dependent manner, while wild-type­ A431 cells were not samples was performed using the t-test­ of means. The killed (figure 2A). Similarly, HCC cells expressing endoge- http://jitc.bmj.com/ expression of GPC3 in human tissues and serum were nous GPC3 responded well to the treatment of hYP7 CAR-­T compared using the Kruskal-­Wallis test and Mann-­Whitney cells and 32A9 CAR-­T cells but not to the parallel treatment U test, respectively. Two-way­ analysis of variance and the of anti-­CD19 CAR-­T cells (figure 2B). When cocultured with log-­rank test were used for tumor growth curve and mouse HCC cells, CAR-­T cells targeting GPC3 showed increases in survival curve analysis. A P<0.05 was considered statistically the release of IL-2, interferon-γ and tumor necrosis factor-α significant. compared with mock-­transduced T cells (figure 2C,D), on September 30, 2021 by guest. Protected copyright. indicating efficient activation of hYP7 CAR-T­ cells and 32A9 CAR-­T cells in vitro. Moreover, we treated immunodeficient RESULTS mice carrying Hep3B xenograft tumors with GPC3-specific­ Construction of CAR-T cells targeting GPC3 in HCC CAR-­T cells or CD19-specific­ CAR-T­ cells. Most of the As GPC3 has been reported to be an HCC-specific­ regu- tumors in mice receiving GPC3-­specific CAR-­T cells showed lator that promotes tumor growth, we first examined its complete remission (CR), whereas those receiving CD19-­ expression in both HCC tumor tissues and HCC cell lines. specific CAR-T­ cells exhibited rapid tumor growth that was Immunohistochemistry staining showed that GPC3 was as fast as that of PBS-treated­ tumors (figure 2E,F). Taken expressed on the membrane and in the cytoplasm of HCC together, the results indicated that hYP7 CAR-T­ cells and tumor tissues but not in peritumor liver tissues or non-­ 32A9 CAR-T­ cells were feasible and effective HCC immuno- tumor liver tissues (figure 1A). Huh7, Hep3B and HepG2, therapies against GPC3 in vitro and in vivo. the three widely used liver cancer cell lines, all showed sGPC3 inhibited HCC cell proliferation and blocked the strong expression of GPC3 in whole-cell­ lysates and on the activation of Wnt signaling induced by GPC3 cell membrane (online supplemental figure S1). We next To investigate the role of sGPC3, we first detected knocked out GPC3 expression in Hep3B cells via a CRIS- the serum concentration of sGPC3 in HCC patients. PR-­Cas9 strategy21 and then s.c. inoculated them into nude Compared with the overall low expression of sGPC3 mice. In contrast to the rapid tumor growth observed in in non-­tumor patients, 6 of 30 HCC patients exhibited the wild-­type Hep3B group, nude mice inoculated with significantly elevated expression of sGPC3 (15–65 ng/mL)

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 3 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 1 Construction of two CAR-­T cells targeting GPC3. (A) IHC staining of GPC3 in human HCC tumor tissues (n=26), peritumor tissues (n=25) and non-­tumor liver tissues (n=9). The staining intensity index was recorded as the total staining intensity per unit of tissue area. Scale bars: 4×, 500 µm; 20×, 100 µm. **P<0.01 (Kruskal-­Wallis test). (B) In vivo tumor growth comparison of Hep3B cells and Hep3B-­GPC3 KO cells. Five million Hep3B cells and Hep3B-­GPC3 KO cells were subcutaneously inoculated into BALB/c nu/nu mice. Each group: n=7. Western blot showed the KO efficiency of GPC3. Values represent the mean±SE ****P<0.0001 (two-­way ANOVA). (C) Schematic Cartoon of GPC3 and anti-­GPC3 monoclonal antibodies (blue, hYP7; red, 32A9). (D) Flow cytometry to detect the binding activity of hYP7 and 32A9 on A431-GPC3­ cells. Gray peak: isotype control. (E) Schematic diagrams of anti-­GPC3 cars designed in the second-­generation 4-­1BB format. Anti-­CD19 CAR based on the FMC63 antibody was used as a control CAR. (F) The representative transduction efficiency of anti-GPC3­ CARs and anti-­CD19 CAR in T cells with the lentivirus system. GFP-positive­ cells were gated as CAR-­T cells. ANOVA, analysis of variance; CAR, chimeric antigen receptor; GFP, green fluorescent protein; GPC3, glypican-3; HCC, hepatocellular carcinoma; hYP7, humanized YP7; IHC, immunohistochemistry; KO, knockout; WT, wild type.

(figure 3A). We then constructed sGPC3-overexpressing­ culture medium. Although we did not directly observe Hep3B cells and Huh7 cells by transfecting a truncated the expression of sGPC3 in the HCC supernatant, Hep3B GPC3 lacking the GPI attachment domain. Both cell lines cells did secrete sGPC3, which could be detected after secreted a large quantity of glycosylated sGPC3 into the being concentrated in an immunoprecipitation assay

4 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 2 Antitumor activity of GPC3-specific­ CAR-­T cells in vitro and in vivo. (A, B) LDH assays to show the cytotoxicity of hYP7 CAR-­T cells and 32A9 CAR-­T cells on A431 cells and A431-­GPC3 cells (A) or native liver cancer cells (B). CAR-­T cells and target cells were cocultured at the indicated E/T ratios for 16 hours. (C, D) ELISA to measure the released cytokines of hYP7 CAR-­T cells (C) and 32A9 CAR-­T cells (D) when targeting HCC cells for 24 hours at an E/T ratio of 1:1. (A–D) experiments were repeated with T cells derived from four donors. Data from one of four repeated experiments are presented as the mean±SD. (E) In vivo antitumor activity of anti-­GPC3 CAR-­T cells. NCG mice with Hep3B-­luciferase xenograft tumors were intravenous injected with two doses of 10 million hYP7 CAR-T­ cells, 32A9 CAR-T­ cells or CD19 CAR-T­ cells when the average tumor size reached approximately 100 mm3. Tumor size was measured via bioluminescence imaging weekly (E) or with a digital caliper every other day (F). The arrow indicates the time point of CAR-T­ cell infusion. Each group: n=7. Values represent the mean±SE *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. (A–E) Two-­tailed Student’s t-test;­ (F) Two-way­ ANOVA. ANOVA, analysis of variance; CAR, chimeric antigen receptor; E/T, effector: target; GFP, green fluorescent protein; GPC3, glypican-3; HCC, hepatocellular carcinoma; hYP7, humanized YP7; IHC, immunohistochemistry; LDH, lactate dehydrogenase; Luc, luciferase; N.D., undetectable; WT, wild type.

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 5 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 3 The inhibitory effect of sGPC3 on HCC cell proliferation and the activation of Wnt signaling in vitro. (A) ELISA to detect the serum concentration of sGPC3 in non-­tumor patients (n=10) and HCC patients (n=30), values represent the mean±SD *P<0.05 (Mann-­Whiteny U test). (B) Western blot to detect the expression of sGPC3 and total GPC3 in sGPC3-over­ expressing HCC cells (left), and immunoprecipitation to compare the expression of sGPC3 in the culture supernatants of Hep3B and Hep3B-sGPC3­ cells (right). (C) Flow cytometric analysis of surface GPC3 expression in sGPC3-expr­ essing HCC cells. Dashed line: isotype control, full line: GPC3 staining. (D) WST-8 assays to detect the proliferation of sGPC3-over­ expressing HCC cells in vitro. (E) TOPFlash reporter assay to detect the mGPC3-­induced activation of Wnt signaling. mGPC3 transfected HEK293 TOPFlash cells were cotransfected with sGPC3, then followed by Wnt3a conditioned medium stimulation. (F) Schematic construct of the recombinant sGPC3-­his protein and its binding activity to anti-GPC3­ monoclonal antibodies. (G) WST-8 assays to detect the cell proliferation of Hep3B cells treated with 50 µg/mL sGPC3-his­ protein for 4 days. (H) TOPFlash reporter assay to detect the mGPC3-­induced activation of Wnt signaling. mGPC3 transfected HEK293 TOPFlash cells were treated with the sGPC3-his­ protein, then followed by Wnt3a conditioned medium stimulation. Data from one of three repeated experiments are presented as the mean±SD *P<0.05; **P<0.01; ***P<0.001 (D, E, G, H) two-­tailed Student’s t-test).­ GPC3, Glypican-3; HCC, hepatocellular carcinoma; IB, immunoblot; IP, immunoprecipitation; mGPC3, membrane GPC3; ns, not significant; sGPC3, shed GPC3.

6 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from (figure 3B). The introduction of sGPC3 did not affect the inhibition was closely correlated with the concentration of expression of cell mGPC3 since sGPC3-overexpressing­ serum sGPC3 (figure 4G). To confirm whether the inhib- and mock-­transfected cells showed equivalent amounts of itory effect was induced by serum sGPC3, we selected two surface GPC3 staining (figure 3C). Next, we compared sGPC3-­positive serum and performed antibody absorp- the cell proliferation rate of mock-­transfected cells and tion. After we added 32A9-­agrose protein A beads into sGPC3-­overexpressing HCC cells. The results showed that the serum to remove the sGPC3, the cytotoxicity of hYP7 sGPC3-­overexpressing cells grew at a significantly lower CAR-­T cells was reversed (figure 4H). These results indi- rate than mock-transfected­ cells, suggesting the inhibitory cated that sGPC3 directly mediated the inhibitory effect effect of sGPC3 on HCC cell proliferation (figure 3D). Wnt of the serum on the cytotoxicity of CAR-T­ cells. signaling is one of the key cancer signaling pathways regu- lating HCC tumor progression, and our previous investi- Hep3B-sGPC3 cells exhibited a worse response to the gation demonstrated a precise mechanism of GPC3 acting treatment of GPC3-specific CAR-T cells in vivo as a Wnt coreceptor.21 We found that the mGPC3-­induced To evaluate the inhibitory effect of sGPC3 on CAR-T­ cell activation of the Wnt-­TOPFlash reporter was suppressed therapy in vivo, we s.c. inoculated Hep3B-Mock­ cells or by the co-­expression of sGPC3 (figure 3E), indicating the Hep3B-­sGPC3 cells into NCG mice. Consistent with our negative regulation on mGPC3-induced­ cancer signaling in vitro results, the tumors in the Hep3B-­sGPC3 group by sGPC3. In addition to sGPC3-­overexpressing cells, we showed delayed tumor initiation compared with those in also purified the recombinant sGPC3 protein fused with the Hep3B-Mock­ group. Hep3B-sGPC3­ cells took longer a hexahistidine tag. The recombinant sGPC3 could be time to grow to same tumor size in vivo (figure 5A). As we recognized by both 32A9 and YP7, indicating its confor- would start CAR-­T cell treatment when mice of two groups mation is similar to native GPC3 (figure 3F). Similarly, were under the same tumor burden, we first evaluated recombinant sGPC3 also blocked the growth of Hep3B whether Hep3B-­sGPC3 tumors exhibited different biolog- cells and inhibited mGPC3-induced­ activation of the Wnt-­ ical features compared with Hep3B-Mock­ tumors under TOPFlash reporter (figure 3G,H). Taken together, these this condition. We did not observe significant difference results indicated that the truncated sGPC3 used here of surface GPC3, β-catenin activation and Ki-67 staining exhibited similar structure identity and biological func- when Hep3B-­Mock tumors and Hep3B-sGPC3­ tumors tion as the native sGPC3, and could conceptually mimic were in similar size (figure 5B-­D). These results might the sGPC3 in our study. help us to exclude the potential fundamental changes of the tumor caused by overexpressing sGPC3. sGPC3 blocked the activation and cytotoxicity of GPC3- When the average tumor volume of each group reached specific CAR-T cells in vitro approximately 180 mm3 (day 16 for the mock group vs Given that sGPC3 is able to compete with mGPC3 for day 21 for the sGPC3 group), the mice were infused with the binding of its ligand during tumor growth, it is also 32A9 CAR-T­ cells intravenously one time and observed possible that anti-­GPC3 CAR-T­ cells can be distracted for 3 weeks (figure 5E). Our results showed that both the http://jitc.bmj.com/ by sGPC3 around tumor cells during the recognition of Hep3B-­Mock tumors and Hep3B-­sGPC3 tumors in mice GPC3-positive­ cells, and the activation and cytotoxicity of treated with CAR-T­ cells started to shrink (below the initial anti-GPC3­ CAR-T­ cells are therefore hampered to some size) within a week. At the end of week three, when the extent. As we predicted, hYP7 CAR-T­ cells and 32A9 tumors in the PBS groups reached the endpoint, the sizes CAR-T­ cells showed dramatically impaired cytotoxicity of most of the tumors in the Hep3B-Mock­ group treated

(figure 4A), deficient cytokine release (figure 4C and with CAR-­T cells were still below the limit of detection, on September 30, 2021 by guest. Protected copyright. online supplemental figure S3A) when targeting sGPC3-­ whereas those in the Hep3B-sGPC3­ group rapidly grew overexpressing HCC cells compared with HCC-Mock­ cells. back after week 2. The CR ratio of the mice was 13/15 Similar impaired effects were also observed when recom- in the Hep3B-­Mock group but only 3/15 in the Hep3B-­ binant GPC3 was added to the wells containing GPC3-­ sGPC3 group at the end of week three after CAR-T­ cell specific CAR-­T cell and HCC cocultures (figure 4B,D and treatment (figure 5F,G). Therefore, 32A9 CAR-­T cells online supplemental figure S3B,C). showed less efficacy in mice with Hep3B-sGPC3­ tumors than in those with Hep3B-­Mock tumors. HCC patient-derived sGPC3 inhibited the cytotoxicity of CAR-T Next, we assessed the role of sGPC3 in CAR-T­ cytotox- cells icity against larger tumors in vivo. The mice were intra- To confirm the blocking effect of patient-derived­ sGPC3 venously injected with two doses (the first contained on CAR-­T cells, we freshly collected eight HCC patients’ 5×106 T cells; the second, 1×107 T cells) of GPC3-­specific serum to perform ex vivo examination. Totally four CAR-­T cells weekly when the tumor volume reached samples had detectable sGPC3 (50% of total samples) approximately 400 mm3 (figure 6A and online supple- in which two samples exhibited high level (25% of total mental figure S4). Under 32A9 CAR-T­ cell treatment, samples) (figure 4E). We then treated hYP7 CAR-T­ and approximately half of the tumors in the Hep3B-Mock­ Hep3B cocultured cells with the patients’ serum. We group started to shrink after the second infusion, and the found that the serum with high expression level of sGPC3 survival of the mice was significantly extended; however, exhibited reduced cytotoxicity (figure 4F), and such the tumors in the Hep3B-sGPC3­ group did not exhibit

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 7 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 4 sGPC3 inhibited the activation and cytotoxicity of GPC3-­specific CAR-­T cells in vitro. (A, B) LDH assays to detect the cytotoxicity of GPC3-­specific CAR-­T cells targeting Hep3B-­Mock cells and Hep3B-­sGPC3 cells at the indicated E/T ratio for 16 hours (A) or targeting Hep3B cells in the presence of recombinant sGPC3-his­ protein at an E/T ratio of 2:1 for 16 hours (B). (C, D) The detection of the cytokine release of hYP7 CAR-­T cells and 32A9 CAR-­T cells targeting Hep3B-­sGPC3 cells for 16 hours at the indicated E/T ratio (C) or targeting Hep3B cells in the presence of recombinant sGPC3-his­ protein at an E/T ratio of 2:1 for 16 hours (D). (E) ELISA to detect the serum concentration of sGPC3 in HCC patients. (F) LDH assays to detect the cytotoxicity of hYP7 CAR-T­ cells targeting Hep3B cells at E/T ratio of 1:1 for 12 hours in the presence of 5% serum from HCC patients. (G) Negative correlation between serum sGPC3 level and cytotoxicity of hYP7 CAR-T­ cells (Pearson correlation). (H) Cytotoxicity of hYP7 CAR-T­ cells treated with HCC serum samples which were pre-­cleaned by 1 µg/mL anti-GPC3­ antibody (32A9) or human IgG. (A–D) Experiments were repeated with T cells derived from three donors. Data from one of three repeated experiments are presented as the mean±SD *P<0.05; **P<0.01; ***P<0.001 (two-tailed­ Student’s t-test).­ CAR, chimeric antigen receptor; E/T, effector: target; GFP, green fluorescent protein; GPC3, Glypican-3; HCC, hepatocellular carcinoma; hYP7, humanized YP7; LDH, lactate dehydrogenase; N.D., undetectable; ns, not significant; sGPC3, shed GPC3.

8 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 5 sGPC3 caused an earlier tumor relapse after 32A9 CAR-­T cell treatment in vivo. (A) Tumor growth of NCG mice inoculated with 5 million Hep3B-Mock­ or Hep3B-­sGPC3 cells subcutaneously. Each group: n=6. Values represent the mean±SE ****P<0.0001 (two-way­ ANOVA). (B) Flow cytometry analysis of surface GPC3 on tumor cells isolated from Hep3B-Mock­ or Hep3B-­sGPC3 xenograft in NCG mice. Each group: n=3. (C) Detection of total and active β-catenin in Hep3B-Mock­ cells or Hep3B-sGPC3­ tumor xenograft by Western blotting. Low: low tumor burden (~250 mm3), high: high tumor burden (~800 mm3). (D) IHC staining of Ki-67 expression of Hep3B-­Mock or Hep3B-­sGPC3 xenograft. Scale bar:100 µm and n=3 in each group. (E) Schematic diagram of CAR-T­ cell treatment. NCG mice were subcutaneously inoculated with luciferase-expr­ essing Hep3B-­ Mock and Hep3B-sGPC3­ cells. When the average tumor size of each group reached ~180 mm3, 10 million 32A9 CAR-­T cells were infused into mice by intravenous injection. (F) Tumor volume was measured with an electronic caliper every other day. The arrow indicates the time point of CAR-T­ cell infusion. Values represent the mean±SE **P<0.01 (two-­way ANOVA). (G) Weekly bioluminescence imaging of mice after CAR-T­ cell infusion. Each group: n=15. ANOVA, analysis of variance; CAR, chimeric antigen receptor; CR, complete remission; IHC, immunohistochemistry; Luc, luciferase; ns, not significant; sGPC3, shed Glypican-3.

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 9 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 6 sGPC3 induced a worse response to GPC3-­specific CAR-­T cell treatments when mice were under a greater tumor burden. NCG mice were inoculated with luciferase-­expressing Hep3B-­Mock and Hep3B-­sGPC3 cells subcutaneously. When the average tumor size of each group reached ~400 mm3, hYP7 CAR-­T cells and 32A9 CAR-­T cells were infused into mice by intravenous injection. (A) Schematic diagram of CAR-­T cell treatments. (B) Individual tumor growth curve of each group. The arrow indicates the time point of CAR-­T cell infusion. (C, D) Kaplan-­Meier survival curve of the 32A9 CAR-­T cell-­treated groups (C) or hYP7 CAR-T­ cell-­treated groups (D). Mock T cell-­treated group: n=8, 32A9 CAR-T­ cell-tr­ eated groups: n=10 and hYP7 CAR-T­ cell-­treated groups: n=10. ***P<0.001; ****P<0.001 (log-­rank test). CAR, chimeric antigen receptor; hYP7, humanized YP7; ns, not significant; sGPC3, shed Glypican-3.

10 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from obvious shrinkage, and the survival of the mice was not infiltrated immune cells. These components form a rela- considerably improved (ffigure 6B,C). The hYP7 CAR-­T tively isolated but complicated microenvironment which cells exerted stronger cytotoxicity in mice than the 32A9 constitutes a major barrier to the development of an CAR-T­ cells. In response to hYP7-CAR-­ ­T cell treatment, effective strategy for curing solid tumors.37–39 To pursue which extended the survival of mice significantly, the better efficacy, more attention is focused on improving tumors in the Hep3B-Mock­ group exhibited shrinkage poor tumor penetration and/or overcoming the inac- relatively early and grew back gradually, whereas the tivation of immune cells in solid tumors.40 41 However, tumors in the Hep3B-­sGPC3 group were inhibited after whether the tumor antigen itself is involved in the insuf- the second injection and recovered rapidly (figure 6B ficient response of solid tumors to antitumor treatments and D). is still obscure. In general, these results indicated that mice with In the current study, we showed that in addition to sGPC3-­overexpressing Hep3B tumors exhibited a worse GPC3-­mediated cancer signaling and tumor growth, the response to treatment with GPC3-­specific CAR-T­ cells. efficacy of GPC3-­specific CAR-­T cells is also blocked by sGPC3 in vitro and in vivo. This competitive inhibition is sGPC3 competes with mGPC3 for the binding of CAR-T cells consistent with the universal agonistic effect of shed cell-­ To investigate the possible mechanism of the blocking surface proteoglycan on their corresponding membrane effect on CAR-T­ cells mediated by sGPC3, we first incu- proteoglycans. Cell-­surface proteoglycans include two bated recombinant sGPC3 with GPC3-specific­ CAR-­T major subfamilies: syndecans and glypicans.42 43 In contrast cells. Both 32A9 CAR-T­ cells and hYP7 CAR-T­ cells could to the well-­described proteolytic cleavage of syndecans specifically bind to recombinant sGPC3 (figure 7A). To by matrix metalloproteinase, disintegrin and metallo- determine whether sGPC3 could induce the activation proteinase domain-containing­ proteins (ADAMs), and of CAR-­T cells, we treated GPC3-specific­ CAR-­T cells ADAMs with thrombospondin domains,42 there is limited with the culture supernatant from Hep3B-sGPC3­ cells or available evidence demonstrating how the shedding of recombinant GPC3. The cultured Hep3B-sGPC3­ superna- glypicans occurs. Notum, a palmitoleoyl-protein­ carbox- tant did not trigger GPC3-specific­ CAR-T­ cells to express ylesterase, has been reported to induce the release of IL-2 (figure 7B). In addition, we found that free recombi- GPI-anchored­ proteins, including glypicans44; ADAM17 nant GPC3, in either monomeric (sGPC3-his)­ or dimeric is regarded as the specific sheddase cleaving glypican-1.45 format (sGPC3-Fc­ mut), could not induce effective secre- We overexpressed or purified truncated GPC3 lacking tion of IL-2 (figure 7C and online supplemental figure the GPI-anchoring­ domain to mimic the native sGPC3. S5). Therefore, sGPC3 bound to GPC3-­specific CAR-­T Although this truncated form of GPC3 was evaluated to cells normally but failed to induce the efficient activation have similar structure identity and biological function to of CAR-­T cells. Interestingly, we found that sGPC3 exhib- the native GPC3, it is still necessary to identify the detailed ited strong activating effect on CAR-T­ cells when coated shedding process of GPC3, including the cleavage site, on the plate (online supplemental figure S5). These GPC3-­specific sheddase and the leading cause of the http://jitc.bmj.com/ phenomena were consistent with the previous study shedding process of GPC3 in HCC. demonstrating that soluble antigen would not trigger the We constructed sGPC3-overexpressing­ HCC cells to activation of CAR-T­ cells unless they are in dimerization investigate the influence of sGPC3 on CAR-T­ cell treat- 36 or crosslinking format. ment in vitro and in vivo. Under similar tumor burden, We further evaluated whether recombinant sGPC3 sGPC3-­overexpressing-­HCC tumor and Mock-­HCC or Hep3B-­sGPC3 cultured supernatant would affect the tumor exhibited similar levels of surface GPC3 expres- on September 30, 2021 by guest. Protected copyright. recognition of CAR-­T cells and target cells. We found sion, cell proliferation rate and activation of β-­catenin. that the recognition of hYP7 CAR-­T cells and Hep3B cells These phenomena suggested that the impaired anti-­ could be blocked by adding recombinant sGPC3 in a dose-­ tumor activity of CAR-T­ cells observed in our study might dependent manner (figure 7 and online supplemental be mainly caused by the sGPC3-induced­ blocking effect. figure S6). Similar blocking effect was also observed after However, this model might only represent a simplified adding the culture supernatant of Hep3B-­sGPC3 cells shedding process of GPC3. Therefore, it would be neces- (figure 7E). Therefore, these results indicated that sGPC3 sary to further evaluate the sGPC3-induced­ resistance to affect CAR-­T cells as a competitor of mGPC3. CAR-­T treatment in a model with more clinical-­relevance, In summary, our results demonstrated that sGPC3 such as an orthotopic model by implanting patient-­ functions as a dominant-negative­ regulator to compete derived HCC tumors with GPC3 shedding that naturally with mGPC3 for the binding of CAR-T­ cells, leading to occurs. deficient activation and blocked cytotoxicity of GPC3-­ In vitro experiments showed that free sGPC3 did not targeted CAR-­T cells (figure 7F). induce the efficient activation of CAR-T­ cells. Accord- ingly, previous reports showed that monomeric forms of CAR ligands, such as mesothelin (MSLN), cannot DISCUSSION trigger CAR signaling.46 Moreover, others also reported Solid tumors are composed of multiple components, that anti-GFP­ CAR-­T cells and anti-transforming­ growth including tumor cells, stroma cells, blood vessels, and factor-β CAR-­T cells could be activated only when soluble

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 11 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from http://jitc.bmj.com/ on September 30, 2021 by guest. Protected copyright.

Figure 7 sGPC3 blocked GPC3-­specific CAR-­T cells from binding membrane GPC3 positive tumor cells. (A) Flow cytometry to detect the binding activity of the sGPC3- his protein on GPC3-­specific CAR-­T cells. CAR-­T cells were incubated with 5 μg/ ml sGPC3-­his for detection. ELISA was performed to measure the secretion of IL-2 in the culture supernatants of CAR-T­ cells treated with the culture supernatant of Hep3B-­sGPC3 cells (B) or sGPC3-his­ protein (C) for 16 hours. CAR-­T cells cocultured with Hep3B cells were used as positive control. (D, E) Flow cytometry to detect the binding of hYP7 CAR-T­ cells and mCherry positive Hep3B cells. Cells were incubated on ice at an E/T ratio of 5:1 for 1 hour in the presence of sGPC3-his­ protein (D) or culture supernatant of Hep3B-­sGPC3 cells (E). Experiments were repeated with T cells derived from three donors. Data from one of three repeated experiments are presented as the mean±SD. **P<0.01; ***P<0.001. (F) Proposed working model of sGPC3 in GPC3- specific CAR-T­ cell therapy. CAR, chimeric antigen receptor; hYP7, humanized YP7; IL-2, interleukin; MFI, mean florescence intensity; N.D., undetectable; ns, not significant; sGPC3, shed glypican-3; sup., supernatant.

12 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from antigen were dimerized or crosslinked.36 In our study, we Contributors LS, FG, ZG, LA, NL, SM and MJ performed the experiments; LS, observed that the crosslinked sGPC3 triggered the activa- FG, TD, SJ, NL, PL and WG analyzed the data; NL constructed the GPC3 (hYP7) CAR plasmid; BS and SJ provided the clinical samples; TD, MH and WG designed tion of CAR-­T cells in vitro (online supplemental figure and conducted the research; TD and WG wrote the manuscript; MH revised the S5). This observation indicates that the effect of sGPC3 manuscript. All authors read and approved the final manuscript. on CAR-­T cells would be more complicated if sGPC3 was Funding This research was supported by the National Natural Science Foundation crosslinked in HCC tumor locally through its HS chains of China (No. 81773260; No. 81972284) and the National Natural Science or other extracellular matrix molecules. Therefore, it Foundation Youth Project of Jiangsu, China (No. BK20171047). The development of the YP7 and humanized YP7 (hYP7) antibodies was supported by the Intramural would be interesting to examine this point in orthotopic Research Program of the National Institutes of Health (NIH), National Cancer mouse model or a mouse HCC tumor model. Besides Institute (NCI) (Z01 BC010891 and ZIA BC010891 to MH). activation, more detailed issues such as T cell exhaustion Competing interests No, there are no competing interests. should also be examined to evaluate the effect of cross- Patient consent for publication Not required. linked shed antigen. Ethics approval All mice received humane care according to the criteria outlined Our study provides a proof-of-­ ­concept example in the Guide for the Care and Use of Laboratory Animals prepared by the National showing that the antigen shedding might cause worse Academy of Sciences and published by the National Institutes of Health. All mice response to immunotherapy in HCC. A recent study on were housed and treated under the protocol approved by the Ethics Committee of secreted Programmed cell death-Ligand­ 1 (sPD-L1)­ vari- the Animal Core Facility of Nanjing Medical University. Analysis with patient samples was approved by the Institutional Review Boards of the participating hospitals. ants showed that sPD-­L1 functioned as a decoy of a PD-L1­ Provenance and peer review Not commissioned; externally peer reviewed. blockade antibody.29 This inhibitory trend of sPD-L1­ is consistent with our current observations to some extent. Data availability statement All data relevant to the study are included in the article or uploaded as online supplemental information. MSLN presents a substantial amount of shedding in many Supplemental material This content has been supplied by the author(s). It has tumor types. However, MSLN shedding seems to exert no not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been significant blocking effect on MSLN-specific­ CAR-­T cells peer-­reviewed. Any opinions or recommendations discussed are solely those in vitro.46 Thus, whether the sGPC3-induced­ inhibitory of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and effect on CAR-­T cells observed here is only an antigen-­ responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability specific phenomenon or represents a universal immune of the translations (including but not limited to local regulations, clinical guidelines, escape mechanism in solid tumors requires further terminology, drug names and drug dosages), and is not responsible for any error investigation. and/or omissions arising from translation and adaptation or otherwise. In conclusion, we have generated novel anti-­GPC3 Open access This is an open access article distributed in accordance with the CAR-T­ cells for the treatment of HCC and demonstrated Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially­ , that sGPC3 attenuates the antitumor activities of CAR-T­ and license their derivative works on different terms, provided the original work is cells in vitro and in vivo, possibly by competing with properly cited, appropriate credit is given, any changes made indicated, and the use membrane-­bound GPC3 to bind CAR-­T cells, indicating is non-­commercial. See http://creativecommons.​ ​org/licenses/​ ​by-nc/​ ​4.0/.​ that sGPC3 may serve as a meaningful indicator for the ORCID iD http://jitc.bmj.com/ prognosis of CAR-T­ therapy. Our results provide a new Wei Gao http://orcid.​ ​org/0000-​ ​0002-7889-​ ​1662 understanding of the mechanisms of tumor immune escape in HCC patients after receiving immunotherapies targeting GPC3. REFERENCES 1 Llovet JM, Zucman-­Rossi J, Pikarsky E, et al. Hepatocellular Author affiliations carcinoma. Nat Rev Dis Primers 2016;2:16018. on September 30, 2021 by guest. Protected copyright. 1Key Laboratory of Human Functional Genomics of Jiangsu Province, National 2 Kulik L, El-­Serag HB. Epidemiology and management of Health Commission Key Laboratory of Antibody Techniques, School of Basic Medical hepatocellular carcinoma. Gastroenterology 2019;156:477–91. Sciences, Nanjing Medical University, Nanjing, Jiangsu, China 3 Villanueva A. Hepatocellular carcinoma. N Engl J Med 2019;380:1450–62. 2Department of Nephrology, The Affiliated BenQ Hospital of Nanjing Medical 4 Greten TF, Lai CW, Li G, et al. Targeted and immune-­based therapies University, Nanjing, Jiangsu, China for hepatocellular carcinoma. Gastroenterology 2019;156:510–24. 3School of Chemistry and Molecular Biosciences, The University of Queensland - 5 Hou J, Zhang H, Sun B, et al. The immunobiology of hepatocellular Saint Lucia Campus, Saint Lucia, Queensland, Australia carcinoma in humans and mice: basic concepts and therapeutic 4Department of Biotherapy, Nanjing Jinling Hospital, Nanjing, Jiangsu, China implications. J Hepatol 2020;72:167–82. 6 Filmus J, Capurro M. Glypican-3: a marker and a therapeutic target 5Laboratory of Molecular Biology, Center for Cancer Research, National Cancer in hepatocellular carcinoma. Febs J 2013;280:2471–6. Institute, National Institutes of Health, Bethesda, Maryland, USA 7 Feng M, Ho M. Glypican-3 antibodies: a new therapeutic target for 6 Department of Medical Oncology, Wuxi People’s Hospital Affiliated to Nanjing liver cancer. FEBS Lett 2014;588:377–82. Medical University, Wuxi, Jiangsu, China 8 Shimizu Y, Suzuki T, Yoshikawa T, et al. Next-­Generation cancer 7Department of Hepatobiliary Surgery, The Affiliated Drum Tower Hospital of Nanjing immunotherapy targeting glypican-3. Front Oncol 2019;9:248. 9 Nishida T, Kataoka H. Glypican 3-T­ argeted therapy in hepatocellular University Medical School, Nanjing, Jiangsu, China carcinoma. Cancers 2019;11:1339. 10 Gao W, Tang Z, Zhang Y-­F, et al. Immunotoxin targeting glypican-3 Acknowledgements We thank Yen Phung and Yi-­Fan Zhang (National Cancer regresses liver cancer via dual inhibition of Wnt signalling and protein Institute or NCI) for isolating and humanizing the YP7 antibody, and Dr. Mei Zhang synthesis. Nat Commun 2015;6:6536. (First Affiliated Hospital of Nanjing Medical University) for optimizing the CAR-T­ 11 Fu Y, Urban DJ, Nani RR, et al. Glypican-3-Specific­ antibody drug conjugates targeting hepatocellular carcinoma. Hepatology cell culture protocol. The National Cancer Institute (NCI) holds patent rights to the 2019;70:563–76. anti-­GPC3 YP7 antibody in many jurisdictions, including the USA, China, Japan, 12 Jiang Z, Jiang X, Chen S, et al. Anti-GPC3-­ ­CAR T cells suppress the South Korea, Singapore, and Europe. Anyone interested in licensing the YP7 growth of tumor cells in patient-derived­ xenografts of hepatocellular antibody can contact Dr. Mitchell Ho (NCI) for additional information. carcinoma. Front Immunol 2016;7:690.

Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875 13 Open access J Immunother Cancer: first published as 10.1136/jitc-2020-001875 on 8 April 2021. Downloaded from 13 Liu Y, Di S, Shi B, et al. Armored inducible expression of IL-12 30 Phung Y, Gao W, Man Y-­G, et al. High-­Affinity monoclonal antibodies enhances antitumor activity of Glypican-3-T­ argeted chimeric antigen to cell surface tumor antigen glypican-3 generated through a Receptor-­Engineered T cells in hepatocellular carcinoma. J Immunol combination of peptide immunization and flow cytometry screening. 2019;203:198–207. MAbs 2012;4:592–9. 14 Liu G, Rui W, Zheng H, et al. CXCR2-­modified CAR-­T cells have 31 Zhang Y-F­ , Ho M. Humanization of high-af­ finity antibodies targeting enhanced trafficking ability that improves treatment of hepatocellular glypican-3 in hepatocellular carcinoma. Sci Rep 2016;6:33878. carcinoma. Eur J Immunol 2020;50:712–24. 32 Liu X, Gao F, Jiang L, et al. 32A9, a novel human antibody for 15 Li D, Li N, Zhang Y-­F, et al. Persistent polyfunctional chimeric designing an immunotoxin and CAR-­T cells against glypican-3 in antigen receptor T cells that target glypican 3 eliminate hepatocellular carcinoma. J Transl Med 2020;18:295. orthotopic hepatocellular carcinomas in mice. Gastroenterology 33 Kiyoshi M, Caaveiro JMM, Kawai T, et al. Structural basis for 2020;158:2250–65. binding of human IgG1 to its high-­affinity human receptor FcγRI. Nat 16 Shi D, Shi Y, Kaseb AO, et al. Chimeric antigen Receptor-­Glypican-3 Commun 2015;6:6866. T-­cell therapy for advanced hepatocellular carcinoma: results of 34 Wang X, Mathieu M, Brezski RJ. Igg Fc engineering to modulate phase I trials. Clin Cancer Res 2020;26:3979–89. antibody effector functions. Protein Cell 2018;9:63–73. 17 Capurro MI, Xiang Y-­Y, Lobe C, et al. Glypican-3 promotes the 35 Feng M, Gao W, Wang R, et al. Therapeutically targeting growth of hepatocellular carcinoma by stimulating canonical Wnt glypican-3 via a conformation-­specific single-­domain antibody signaling. Cancer Res 2005;65:6245–54. in hepatocellular carcinoma. Proc Natl Acad Sci U S A 18 Gao W, Ho M. The role of glypican-3 in regulating Wnt in 2013;110:E1083–91. hepatocellular carcinomas. Cancer Rep 2011;1:14–19. 36 Chang ZL, Lorenzini MH, Chen X, et al. Rewiring T-­cell responses 19 Kolluri A, Ho M. The role of glypican-3 in regulating Wnt, YAP, and to soluble factors with chimeric antigen receptors. Nat Chem Biol hedgehog in liver cancer. Front Oncol 2019;9:708. 2018;14:317–24. 20 Li N, Gao W, Zhang Y-F­ , et al. Glypicans as cancer therapeutic 37 Reina-­Campos M, Moscat J, Diaz-­Meco M. Metabolism shapes the targets. Trends Cancer 2018;4:741–54. tumor microenvironment. Curr Opin Cell Biol 2017;48:47–53. 21 Li N, Wei L, Liu X, et al. A Frizzled-­Like cysteine-­rich domain in 38 Wellenstein MD, de Visser KE. Cancer-­Cell-Intrinsic­ mechanisms glypican-3 mediates Wnt binding and regulates hepatocellular shaping the tumor immune landscape. Immunity 2018;48:399–416. carcinoma tumor growth in mice. Hepatology 2019;70:1231–45. 39 Nguyen KB, Spranger S. Modulation of the immune 22 Kailemia MJ, Park D, Lebrilla CB. Glycans and glycoproteins microenvironment by tumor-­intrinsic oncogenic signaling. J Cell Biol as specific biomarkers for cancer. Anal Bioanal Chem 2020;219:e201908224. 2017;409:395–410. 40 Binnewies M, Roberts EW, Kersten K, et al. Understanding the tumor 23 Capurro M, Wanless IR, Sherman M, et al. Glypican-3: a novel serum and histochemical marker for hepatocellular carcinoma. immune microenvironment (time) for effective therapy. Nat Med Gastroenterology 2003;125:89–97. 2018;24:541–50. 24 Xu D, Su C, Sun L, et al. Performance of serum glypican 3 in 41 Badalamenti G, Fanale D, Incorvaia L, et al. Role of tumor-infiltrating­ diagnosis of hepatocellular carcinoma: a meta-­analysis. Ann Hepatol lymphocytes in patients with solid tumors: can a drop dig a stone? 2019;18:58–67. Cell Immunol 2019;343:103753. 25 Liu S, Wang M, Zheng C, et al. Diagnostic value of serum glypican-3 42 Theocharis AD, Skandalis SS, Tzanakakis GN, et al. Proteoglycans in alone and in combination with AFP as an aid in the diagnosis of liver health and disease: novel roles for proteoglycans in malignancy and cancer. Clin Biochem 2020;79:54–60. their pharmacological targeting. Febs J 2010;277:3904–23. 26 Zittermann SI, Capurro MI, Shi W, et al. Soluble glypican 3 inhibits 43 Gondelaud F, Ricard-­Blum S. Structures and interactions of the growth of hepatocellular carcinoma in vitro and in vivo. Int J syndecans. Febs J 2019;286:2994–3007. Cancer 2010;126:1291–301. 44 Traister A, Shi W, Filmus J. Mammalian notum induces the release 27 Feng M, Kim H, Phung Y, et al. Recombinant soluble glypican 3 of glypicans and other GPI-anchor­ ed proteins from the cell surface. protein inhibits the growth of hepatocellular carcinoma in vitro. Int J Biochem J 2008;410:503–11. Cancer 2011;128:2246–7. 45 Kawahara R, Granato DC, Yokoo S, et al. Mass spectrometry-based­ 28 Kulasingam V, Diamandis EP. Strategies for discovering novel cancer proteomics revealed glypican-1 as a novel ADAM17 substrate. J biomarkers through utilization of emerging technologies. Nat Clin Proteomics 2017;151:53–65. Pract Oncol 2008;5:588–99. 46 Lanitis E, Poussin M, Hagemann IS, et al. Redirected antitumor 29 Gong B, Kiyotani K, Sakata S, et al. Secreted PD-L1­ variants mediate activity of primary human lymphocytes transduced with a http://jitc.bmj.com/ resistance to PD-L1­ blockade therapy in non-small­ cell lung cancer. J fully human anti-­mesothelin chimeric receptor. Mol Ther Exp Med 2019;216:982–1000. 2012;20:633–43. on September 30, 2021 by guest. Protected copyright.

14 Sun L, et al. J Immunother Cancer 2021;9:e001875. doi:10.1136/jitc-2020-001875